radio interface technologies for cooperative transmission

3202IEICE TRANS. COMMUN., VOL.E94–B, NO.12 DECEMBER 2011

INVITED PAPER Special Section on Cooperative Communications for Cellular Networks

Radio Interface Technologies for Cooperative Transmission in3GPP LTE-Advanced

Tetsushi ABE†, Yoshihisa KISHIYAMA†a), Yoshikazu KAKURA††, and Daichi IMAMURA†††, Members

SUMMARY This paper presents an overview of radio interface tech-nologies for cooperative transmission in 3GPP LTE-Advanced, i.e., coor-dinated multi-point (CoMP) transmission, enhanced inter-cell interferencecoordination (eICIC) for heterogeneous deployments, and relay transmis-sion techniques. This paper covers not only the technical components inthe 3GPP specifications that have already been released, but also those thatwere discussed in the Study Item phase of LTE-Advanced, and those thatare currently being discussed in 3GPP for potential specification in futureLTE releases.key words: LTE-Advanced, CoMP, eICIC, relay

1. Introduction

After enthusiastic efforts for standardization in the 3rd Gen-eration Partnership Project (3GPP) and development, a com-mercial Long-Term Evolution (LTE) service was launchedin Japan in December 2010. LTE provides full IP packet-based radio access with low latency and adopts intra-cellorthogonal multiple access schemes such as orthogonalfrequency division multiple access (OFDMA) and single-carrier frequency division multiple access (SC-FDMA) inthe downlink and uplink, respectively [1], [2]. Moreover, inthe 3GPP, standardization efforts towards an enhanced LTEradio interface called LTE-Advanced (LTE Release 10 andbeyond) are ongoing. In the beginning of 2011, specifica-tions for LTE Release 10 were finalized to complete sub-mission to the International Telecommunication Union Ra-diocommunication sector (ITU-R), and technical discussionregarding LTE Release 11 was initiated in the 3GPP.

In LTE initial release, i.e., LTE Release 8 specificationsfinalized in 2008, the radio interface was designed basicallyfocusing on traditional cellular deployments associated withtransmission between the base station, called an eNB (en-hanced Node B), and user equipment (UE) as in 3G mobilecommunication systems. In LTE-Advanced, enhanced net-work deployments based on cooperative transmission tech-niques have been actively investigated to achieve highersystem performance and more efficient/flexible networkdeployments. Figure 1 illustrates cooperative transmis-

Manuscript received June 22, 2011.Manuscript revised August 4, 2011.†The authors are with NTT DOCOMO, INC., Yokosuka-shi,

239-8536 Japan.††The author is with NEC Corporation, Kawasaki-shi, 211-8666

Japan.†††The author is with Panasonic Corporation, Yokohama-shi,

224-8539 Japan.a) E-mail: [email protected]

DOI: 10.1587/transcom.E94.B.3202

Fig. 1 Cooperative transmission techniques in LTE-Advanced.

sion techniques in LTE-Advanced, which are discussedin the paper. Coordinated multi-point (CoMP) transmis-sion/reception is based on the transmission techniques usingcooperation among multiple cells or geographically sepa-rated transmission points. The enhanced inter-cell interfer-ence coordination (eICIC) for heterogeneous deployments isa transmission technique that utilizes cooperation betweencells or nodes with different transmission power levels. Therelay transmission techniques utilize cooperation betweeneNB and relay nodes (RNs) or between a UE and RNs toextend the coverage or capacity. In the 3GPP, some radiointerface technologies to support such cooperative transmis-sion techniques have already been specified in LTE Release10, and their further enhancements are being investigatedfor the specification of LTE Release 11 and beyond.

This paper presents an overview of the LTE-Advancedstandardizations and a survey of the related research to ex-plain the radio interface technologies for cooperative trans-mission such as CoMP, eICIC, and relay in LTE-Advanced.In the paper, CoMP, eICIC, and relay topics are explainedin Sects. 2, 3, and 4, respectively. Note that the scope of thepaper includes not only Release 10 specifications, but alsotechnical discussions held during the Study Item (SI) phasefor LTE-Advanced (LTE-Advanced SI) [3] and further ad-vanced techniques that may be specified in future LTE re-leases.

2. Coordinated Multiple-Point Transmission and Re-ception

The evolution from W-CDMA to OFDMA/SC-FDMA inLTE has enabled orthogonal multiple access within a cell.On the other hand, inter-cell interference, due to static singlefrequency reuse, is still a bottleneck for further improvingthe spectrum efficiency. Thus, for LTE-Advanced, CoMP

Copyright c© 2011 The Institute of Electronics, Information and Communication Engineers

ABE et al.: RADIO INTERFACE TECHNOLOGIES FOR COOPERATIVE TRANSMISSION IN 3GPP LTE-ADVANCED3203

transmission and reception techniques have been studied toimprove the spectrum efficiency based on the idea of ex-tending static single frequency reuse to dynamic radio re-source orthogonalization among cells [4]. In the 3GPP, theterm “CoMP” was defined during the LTE-Advanced SI[3], where basic technologies and terminologies related tointer-cell cooperative processing or network multiple-inputmultiple-output (MIMO) technologies [5], [6] were inves-tigated. In LTE Release 10, the specifications related toCoMP are limited to the downlink reference signal (RS) tosupport CoMP transmissions. However, for Release 11, fur-ther study is in progress under the new SI (CoMP SI) tar-geting further specification of CoMP related aspects. Notethat downlink CoMP has been mainly discussed in the 3GPPsince uplink CoMP could be implemented to a large ex-tent without additional specifications for the physical layer.Thus, this paper focuses on the downlink aspects of CoMPtechnologies, which would match the scope of this journal.

2.1 Categorization of CoMP Transmission Schemes

In the LTE-Advanced SI, downlink CoMP transmissiontechnologies are categorized into Coordinated Schedul-ing/Beamforming (CS/CB) and Joint Processing (JP), whereJP is further categorized in to Joint Transmission (JT) andDynamic Cell Selection (DCS). Each transmission technol-ogy may require different standardization effort, and whichone should be supported is under investigation in 3GPP.

Figure 2(a) shows an example of CS/CB, where Physi-cal Downlink Shared Channel (PDSCH) is transmitted onlyfrom the serving cell, whereas scheduling and beamformingare coordinated among cells. For example, as depicted inFig. 2(a), each cell can calculate beamforming weights forits scheduled UEs such that interference to the scheduledUEs in neighboring cells is mitigated. Then, the signal-to-interference and noise power ratio (SINR) of cell-edge UEsis improved and so is the cell-edge user throughput.

Fig. 2 CoMP transmission schemes in LTE-Advanced downlink.

Figure 2(b) shows an example of JT of JP, where thePDSCH is transmitted simultaneously from the serving andnon-serving cells. Precoding weights are calculated suchthat the signals transmitted from multiple cells are coher-ently combined at the UE. Then, the SINR of a cell-edge UEis improved and so is the cell-edge user throughput. Further-more, multi-user (MU)-MIMO techniques can be applied toenable advanced MU-JT, where joint precoding is appliedto the multiple antennas located on multiple cells to per-form orthogonal precoding to the scheduled UEs on thesecells. Theoretically, MU-JP could significantly improve notonly the cell-edge user throughput but also the average cellthroughput. However, MU-JT requires very accurate chan-nel state information (CSI) that needs to be shared amongcells to form orthogonal precoders [7], [8].

Figure 2(c) shows an example of DCS of JP, wherethe PDSCH is transmitted from a dynamically selected cellbased on the instantaneous CSI at the eNB. Note that it isassumed in DCS that the serving cell from which the UEreceives the Physical Downlink Control Channel (PDCCH),is not dynamically changed. Thus, DCS is categorized asJP since the PDSCH can be transmitted from non-servingcells. Furthermore, in DCS, if cells that are not selected forPDSCH transmission are kept muted instead of schedulingother UEs in the cell, the SINRs of cell-edge UEs are signifi-cantly improved, and the achievable cell-edge user through-put becomes closer to that of single-user (SU)-JT [4].

Finally, we would like to mention some CoMP relatedtechniques that have not yet been fully discussed in the3GPP. Coordinated power control among cells, referred toas a cell breathing technique [5], can be considered as anadditional alternative to the 3GPP CS/CB scheme that per-forms coordinated scheduling and beamforming. In partic-ular, in coordinated scheduling such as DCS, some cells aremuted to mitigate inter-cell interference, which causes spec-trum loss, whereas coordinated power control in cell breath-ing has an advantage in that it does not cause spectrum loss.Another cooperative transmission technique that is gainingattention recently is an interference alignment (IA) tech-nique, where coordinated beamforming at the eNB and in-terference suppression at the UE receiver are combined [9].Interference suppression receiver is also being studied in the3GPP as an interference rejection combining (IRC) receiver[10]. A typical IRC receiver performs linear MMSE recep-tion to suppress both intra- and inter-cell interference. Thus,when a UE has only two receiver antennas, it could suppressone dominant interfering signal due to the limited degrees offreedom. However, if IA is applied, multiple cells performcoordinated precoding to align interference to the neighbor-ing cell UEs. Then, for example, a UE with 2 receiver an-tennas can suppress interference from two interfering cells.

2.2 Radio Interface for CoMP Transmission

In LTE Release 10, the downlink RS structure required fordownlink CoMP transmission is specified, and thus CoMPtransmission is possible in a UE transparent manner. Here,


the UE transparent manner means that the UE can apply aunified demodulation scheme irrespective of whether or notCoMP transmission is applied at the eNB. Figure 3 showsthe downlink RS structures specified for LTE Release 10.The downlink control signal transmitted on the PDCCH thatspans up to the first 3 OFDM symbols are demodulatedbased on a cell-specific RS (CRS), and thus is transmittedonly from the serving cells. On the other hand, the PDSCHis demodulated based on a UE-specific demodulation RS(DM-RS), the structure for which is shown in Fig. 3(a). Hy-brid frequency division multiplexing (FDM) and code divi-sion multiplexing (CDM) structures are applied to support amaximum 8 layer MIMO transmission to achieve the peakspectrum efficiency of 30 bps/Hz. CDM is combined withFDM because CDM is beneficial in that it allows the samemapping pattern for the DM-RS irrespective of the numberof the MIMO layers, and avoids transmit power boosting ofthe DM-RS compared to the PDSCH. What is noteworthyhere is that the DM-RS is multiplexed only on physical re-source blocks (PRBs) on which the PDSCH is scheduled,and the same transmission scheme, such as precoding, asthe PDSCH is applied to the DM-RS. Thus, the UE doesnot need to be informed by the eNB what precoding weightis applied to the PDSCH, which enables various eNB im-plementations of the CoMP transmission strategy, such asnull-steering for CS/CB and transmission from non-servingcells for JP in a UE transparent manner.

CSI is required at the eNB to support CoMP transmis-sion. Particularly, CS/CB and MU-JT would require accu-rate CSI over multiple cells, and thus a trade-off betweenperformance gain and the CSI feedback overhead shouldcarefully be considered. Figure 3(b) shows the structure ofthe CSI-RS, which is used solely for CSI measurement. Un-like the DM-RS, which is precoded in the same manner as

Fig. 3 Downlink RS structure in LTE-Advanced.

the PDSCH, the CSI-RS is transmitted from each antennaport for a UE to measure the channel of each antenna port.A hybrid CDM and FDM structure is applied to multiplexmultiple antenna ports. In addition, the density of the CSI-RS is 1 resource element (RE) per antenna port per PRB,and the CSI-RS is periodically transmitted, where the mini-mum duty cycle is 5 msec. Thus, overall the CSI-RS densityis much sparser than that for the DM-RS. This is becausethe CSI-RS is used only for CSI measurement and thus re-quires lower density than the DM-RS, which is used fordata demodulation. What is noteworthy here is that whileLTE Release 10 does not support multi-cell CSI feedback,that is required to support cooperative processing for CoMPtransmissions, i.e., scheduling and beamforming over mul-tiple cell sites [7], [8], CSI-RS is designed to enable inter-cell CSI measurement for further CoMP feedback schemes.One such feature is PDSCH muting that achieves orthogonalCSI-RS transmission between cells, where each cell mutesthe PDSCH REs corresponding to the CSI-RS REs of theneighbor cells. To achieve a high multiplexing capabilityof CSI-RS among cells, multiple CSI-RS patterns are speci-fied per PRB, where Fig. 3(b) illustrates one example of suchpatterns.

2.3 Network Deployment for CoMP Transmission

Coordinated network architecture for CoMP could be basedon inter-eNB distributed coordination or intra-eNB central-ized coordination using remote radio heads (RRHs) [4]. Inrecent years, heterogeneous deployments, where low powernodes to deploy pico and femto cells are overlaid on macrocells, are gaining attention as a solution to offload increas-ing amounts of traffic with limited cost. Thus, the currentCoMP SI in Release 11 studies CoMP transmission for theheterogeneous deployments as shown in Fig. 4(b) as well asthe homogeneous deployments as shown in Fig. 4(a) con-sidering both distributed and centralized coordination. Es-pecially, Fig. 4(b) illustrates a promising network architec-ture for heterogeneous deployments, where pico cells with asmall cell radius are overlaid on the macro cell with a largecell radius, and pico cells are deployed using RRHs that areconnected to the macro eNB so that the central coordinationis applied without latency issues for the coordination amongcells. Furthermore, 3GPP studies two deployment options,where each pico cell has identical cell identification (cell ID)to macro cell, and each pico cell has a distinct cell ID [11].

The intra-eNB central coordination using RRHs, as

Fig. 4 Network deployments assumed in 3GPP CoMP SI.


such, enables more dynamic coordination of radio resourcescompared to the inter-eNB distributed coordination, whichwould provide higher CoMP gain. On the other hand,the RRH deployment requires high capacity optical fibersin backhaul connections, and the complexity of the cen-tral controller is increased as the number of coordinatedcells/RRHs is increased. Thus, the actual number and thesize of coordinated cells would be restricted by the deploy-ment cost. Therefore, one important challenge for the CoMPinvestigation is to establish optimum network architecturesby taking into account the trade-off between deploymentcost and network capacity improvement by using relevantcriteria, e.g., spectrum efficiency per cost.

3. Enhanced Inter-Cell Interference Coordination forHeterogeneous Deployments

Recently, the amount of mobile traffic has been increasingrapidly with the spread of high speed mobile devices suchas smart phones. Therefore, further capacity enhancementin cellular networks is required. The enhancement just byincreasing the system bandwidth and spectrum efficiency isnot enough to meet this demand and the introduction of lowpower eNBs is inevitable. In the 3GPP, pico-eNBs, whichare open to all the subscribers named as Open SubscriberGroup (OSG) and femto-eNBs which are open to the limitedsubscribers named as Closed Subscriber Group (CSG), havebeen defined. In heterogeneous deployments where eNBswith different transmission power levels exist, new inter-cellinterference issues have been found.

This section explains interference scenarios in hetero-geneous deployments, the solutions for which have beenspecified in Release 10 and the issues to be discussed in Re-lease 11.

3.1 Interference Scenarios in Heterogeneous Deployments

In LTE-Advanced SI, interference scenarios in heteroge-neous deployments have been identified [3]. Figure 5 illus-trates interference scenarios in a macro-femto environment.Three scenarios have been identified in LTE-Advanced SI.The first scenario is downlink interference from a femto-eNB to a macro-UE who is not a CSG member of the femto-eNB as shown in (a) of Fig. 5. The second one is uplink in-terference from a macro-UE to a femto-eNB as shown in (b).The third one is downlink interference from a femto-eNB to

Fig. 5 Interference scenarios in macro-femto environment.

a femto-UE who is a CSG member of a different femto-eNBas shown in (c).

Figure 6 illustrates an interference scenario in a macro-pico environment. This scenario assumes downlink interfer-ence from a macro-eNB to a pico-UE at the cell edge whencell selection is based on path loss which is optimum to theuplink channel.

eICIC specifications in Release 10 focus on co-channelinterference in the downlink control channel. It is becauseconventional interference mitigation schemes on data chan-nels such as fractional frequency reuse (FFR) cannot be ap-plied to the control channel due to restrictions in controlsignal mapping, and then the control channel capacity canbecome a bottleneck. For a macro-femto environment, thescenario in (a) of Fig. 5 was assumed. For a macro-pico en-vironment, the downlink interference from the macro-eNBto pico-UE at the cell edge shown in Fig. 6 was assumed.Instead of cell selection based on path loss, that with CellRange Expansion (CRE) [12], [13] was assumed. CRE is atechnique that enhances the traffic offload by giving offset toreceived power from a low power eNB in the cell selectionprocedure. With an increase in the offset value, the possibil-ity that a UE selects a low power eNB increases.

3.2 Inter-Cell Interference Coordination Techniques inHeterogeneous Deployments

In order to mitigate co-channel interference in the downlinkcontrol channel, power setting and time domain solutions[14] have been specified in Release 10 as eICIC techniques.

Power setting is targeted for mitigation of downlink in-terference from a femto-eNB to a macro-UE. In the casethat a femto-eNB is located at the cell edge of a macro-eNB,downlink performance of the macro-UE near the femto-eNBis degraded significantly due to interference from the femto-eNB as the received power from the macro-eNB is small.Therefore, transmission power of the femto-eNB should berestricted to mitigate interference to the macro-UE. In powersetting, a femto-eNB estimates its interference to a macro-UE. Then, the femto-eNB sets its transmission power sothat interference to the macro-UE is sufficiently mitigatedas shown in Fig. 7.

As power setting schemes, three representative exam-ples are shown below.

1. Power setting based on the maximum received

Fig. 6 Interference scenarios in macro-pico environment.


Fig. 7 Interference mitigation by power setting.

power of a macro-eNB at a femto-eNB [15].2. Power setting based on path loss between a femto-

eNB and a macro-UE [16].3. Power setting based on signal to interference power

ratio at a macro-UE [17].

In scheme 1, for example, a femto-eNB sets its trans-mission power by

Ptx = max(min(PM + Poffset, Pmax), Pmin) (1)

where Ptx is the transmission power of a femto-eNB, PM isthe received power at a femto-eNB from macro-eNB, Poffsetis the fixed offset value, Pmax is the maximum transmissionpower of a femto-eNB, and Pmin is the minimum transmis-sion power of a femto-eNB. As PM becomes smaller, thefemto-eNB judges the macro-eNB is located further that isinterference from the femto-eNB to the macro-UE is larger.Then, the femto-eNB sets its transmission power smaller.

In the case that the path loss between a femto-eNB anda macro-UE is large, interference from the femto-eNB tothe macro-UE becomes smaller. However, this aspect is notconsidered in scheme 1 and scheme 1 may set its transmis-sion power unnecessarily low. Therefore, the power settingin scheme 2 takes into account path loss between the femto-eNB and the macro-UE in estimating Poffset in (1), for ex-ample, by

Poffset = PUL tx − PUL interference + P0 (2)

where PUL tx is the transmission power or its estimatedvalue of a macro-UE, PUL interference is the measuredpower of uplink interference at a femto-eNB and P0 is thefixed value.

In scheme 3, for example, a femto-eNB sets its trans-mission power by

Ptx=max(min(SINR MUE+Poffset, Pmax), Pmin) (3)

where SINR MUE is the measured SINR at a macro-UE.This scheme can directly reflect the received signal qualityat the macro-UE. However, the macro-UE cannot report itsmeasurement results directly to the femto-eNB as such aninterface has not yet been defined in 3GPP. The macro-UEcan only report to the macro-eNB. Therefore, a new inter-face between the macro-eNB and femto-eNB to transfer thereport from the macro-UE or a direct reporting scheme fromthe macro-UE to the femto-eNB must be newly specified toachieve this scheme.

The time domain solution mitigates interference from

Fig. 8 Restriction on signal transmission in time domain solution.

Fig. 9 Restriction on signal transmission in MBSFN subframe.

an eNB to another eNB by restricting the signal transmissionfrom the eNB in the time domain.

Figure 8 illustrates an example of interference mitiga-tion with time-domain restrictions on the signal transmis-sion. This figure is an example for a case in which interfer-ence from a femto-eNB to a macro-UE is mitigated. Here,it is assumed that the femto-eNB and the macro-eNB aresynchronized. By restricting signal transmission in specificsubframes at the femto-eNB, interference to the macro-UEcan be mitigated at the timing of the restricted subframes atthe femto-eNB.

The configuration of Multicast/Broadcast over SingleFrequency Network (MBSFN) subframes is a representativescheme to restrict signal transmission at the eNB. The MB-SFN subframe was originally defined to be applied to mul-ticast/broadcast service in LTE Release 8. It is specified inLTE Release 8 that CRSs in the MBSFN subframes are notused for UE measurement and LTE UE recognizes the posi-tion of MBSFN subframes. Therefore, signal transmissioncan be stopped not only for data signals and control signalsbut also for CRS in MBSFN subframes as shown in Fig. 9.It should be noted that MBSFN subframes can be config-ured only at 6 specific subframes in a radio subframe whichconsists of 10 subframes.

An Almost Blank Subframe (ABS) is another kind ofsubframe defined in Release 10 in which restriction to thesignal transmission can be applied. Release 10 UEs recog-nize ABS but prior release UEs do not. Therefore, CRSscannot be stopped in ABS.

In LTE-Advanced, carrier aggregation (CA) whichbundles multiple LTE carriers was specified. In additionto eICIC techniques, which assume co-channel interferencewithin one carrier, a CA-based scheme that assumes multi-


Fig. 10 CA-based interference coordination scheme.

carrier operation has been considered in LTE-Advanced SI.Figure 10 illustrates an example of a CA-based scheme.

It is assumed here that both macro-eNB and pico-eNB uti-lize two carriers and each eNB utilizes different carriers fortransmission of the control signal. In the control signal re-gion of the carrier where each eNB does not transmit a con-trol signal, each eNB does not transmit any signal or trans-mit a signal with low power to mitigate interference to thecontrol signal.

In LTE-Advanced, a control channel in a carrier canindicate resource allocation information of different carriersby defining the carrier indicator field (CIF) and it enablesthe implementation of an interference coordination schemeacross multiple carriers as shown in Fig. 10.

3.3 Remaining Issues in Inter-Cell Interference in Hetero-geneous Deployments

In Release 11, the further enhancement or optimization ofeICIC techniques will be discussed. Release 10 focused onthe interference to downlink control signals. In Release 11,interference mitigation in the uplink control signal and inthe data signal both for the downlink and uplink are to beconsidered.

4. Relay Transmission Techniques

Relaying has been considered for LTE-Advanced as a toolto improve the coverage of high data rates, the cell-edgethroughput, and/or to provide coverage in former un-servedareas. A RN is part of the radio access network and behavessimilar to an eNB from the perspective of a UE. A RN hasa wireless connection to a donor eNB (d-eNB) as illustratedin Fig. 11.

4.1 Relay Techniques Considered in Feasibility Study

Various types of relay techniques can be envisaged. Fea-sibility of those various candidates was studied in LTE-Advanced SI. A simple RF repeater, a so-called amplify-and-forward relays [18], was specified as a part of the Re-lease 8 specifications [19]. However, it is hard to obtain a

Fig. 11 Operation principle of relaying.

substantial improvement in coverage of the higher data ratesdue to their simplicity. Thus, advanced relays were consid-ered as a potential key technology to achieve the require-ments [3].

4.1.1 Relay Node Category

At the beginning of the feasibility study, three types ofdecode-and-forward relays, which decode and re-encodethe received signal before forwarding it, were mainly con-sidered in terms of implemented functionality at the RN.

Layer 1 (L1) relay: An L1 relay is an advanced repeaterequipped with fast Fourier transform (FFT)/inverse FFT(IFFT) as physical layer functionality in order to performfrequency-selective relay and optimum power allocation inthe frequency domain [20].Layer 2 (L2) relay: In addition to the L1 relay, the L2 re-lay utilizes a part of the L2 functions in order to provide,e.g., cooperative transmission [21] and/or cooperative Hy-brid ARQ (HARQ) retransmission between the d-eNB andRN and/or between RNs. Since this type of relay behaves asan integral part of the donor cell it does not have a schedul-ing function nor its own cell ID. A Hybrid of the RF repeaterand L2 relay was also proposed. The hybrid RN amplifiesand forwards initial data, then transmits the re-encoded ini-tial data if re-transmission is required.Layer 3 (L3) relay: In addition to the L2 relay, the L3 re-lay has all functionalities for the radio resource management(RRM) from the perspective of a UE. Thus, a L3 relay hasits own cell ID and controls its own cell, each of which ap-pears to a UE as an individual cell distinct from the donorcell.

One of the most important requirements in the devel-opment of a Release 10 relay is to support legacy UEs (i.e.backward compatibility), that is, the RN behaves as an eNBaccommodating Release 8 UEs. This ensures gradual in-troduction of relay cells into existing networks. Taking thecompatibility, the expected performance improvements andrequired additional specification efforts into account, the L3relay was selected as the relay architecture for Release 10.

4.1.2 Band Assignment between Backhaul and AccessLinks

Since the RN communicates with both d-eNB and UEs be-


longing to the RN (rUEs) avoidance of self-interferencefrom transmission to reception between the backhaul linkand the access link of the relay cell has to be addressed bya reasonable method. Therefore, the time, frequency or spa-tial isolation between the backhaul and access links can beconsidered. In the following, the access link indicates thelink between the RN and rUE unless explicitly stated.

For LTE-Advanced, outband and inband relay opera-tions were defined with respect to the resource isolation be-tween the backhaul and access links [3].- Outband relay: A frequency band assigned to the wire-

less backhaul link is distinct from that assigned to an ac-cess link. Consequently, full duplex relay operation canbe achieved.

- Inband relay: The wireless backhaul link shares thesame frequency band used by the access link within thecells belonging to a d-eNB.

An outband relay can be achieved as an implementation de-tail by reusing Release 8 radio interface (between eNB andUE) for the backhaul link, while the realization of an in-band relay requires careful consideration in terms of back-ward compatibility to LTE Release 8. In Release 10, timedomain separation between backhaul and access links wasintroduced to achieve inband relay operation. The followingsection describes this in detail.

4.2 Relay in LTE-Advanced

As mentioned above, Release 10 relays support basic L3 re-lay functions with backward compatibility to Release 8 UEs[22]. RNs are expected to be stationary, i.e., mobility ofRNs is not assumed. Each RN is connected to a predefinedd-eNB. The number of hops for relaying was limited to two,i.e., one hop as the backhaul link and one hop as the accesslink.

4.2.1 Protocol Architecture

A RN is connected to a d-eNB via the Un interface, and UEsconnect to the RN via the Uu interface as shown in Fig. 12,where the Evolved Packet Core (EPC) is the evolved corenetwork (CN) connected to Release 8 and later releases ofLTE, and S1 is the interface between eNB/RN and EPC [23].To avoid or minimize the impact to the existing UEs andEPC from relay introduction, on the Uu interface, all con-trol plane and user plane protocols are terminated in the RNnamely the RN behaves as an eNB from the UE perspective.For the Un interface, the RN has an S1 interface and behavesas an eNB from the EPC perspective because the eNB actssimilar to an S1 proxy.

4.2.2 Resource Partitioning for Backhaul Link

Time division based resource partitioning at the RN is sup-ported to enable the inband relay operation as shown inFig. 13. Backhaul and access links are allocated to differentsubframes in the same frequency bands for each uplink and

Fig. 12 Relation between nodes and interfaces.

Fig. 13 Resource partitioning and MBSFN subframe usage.

downlink (half duplex). In addition, for TDD, downlink anduplink transmissions on the backhaul link are performed inthe DL and UL subframes at the d-eNB, respectively. Theseaspects allow the sharing of radio resources in the same sub-frame between backhaul links to RNs and access links toUEs belonging to an eNB (mUEs).

To provide the time division resource partitioning be-tween backhaul and access link transmissions in the down-link, i.e., to create “gaps” in the access link transmission,configuring MBSFN subframes in the access link has beenadopted as a way to notify the rUEs (including Release 8UEs) that such subframes contain, e.g., the PDCCHs, CRSsonly in the first one or two OFDM symbols but do not con-tain any PDSCHs or CRSs in the following OFDM sym-bols as illustrated in Fig. 13. Moreover, because Release 8UEs do not carry out the measurement using CRSs in theMBSFN subframes except on the first two OFDM symbols,the gap can be made in the access link without breaking thechannel quality measurement of Release 8 UEs. In the up-link no special treatment is required because the RN justdoes not allocate access link resources for uplink subframeswhere it intends to transmit to the d-eNB.

4.2.3 Downlink Channel Structure for Relay

To support Release 8 UE association to a RN, i.e., back-ward compatibility, PDCCHs that are allocated to the firstone, two, or three OFDM symbols must be transmitted inall subframes on the access link. Therefore, an RN cannotreceive all or a part of the OFDM symbols containing PD-CCH on the backhaul link. Consequently, the Relay-specificPDCCH (R-PDCCH) was newly defined and it is transmit-ted inside the PDSCH region in backhaul subframes.

The RN decodes downlink resource allocation signal-ing (DL grant) and uplink resource allocation signaling (ULgrant) in the R-PDCCH, and then receives the PDSCH and


Fig. 14 R-PDCCH multiplexing and structure.

transmits the physical uplink shared channel (PUSCH) cor-responding to the DL grant and UL grant, respectively. TheDL grants are transmitted only in the first slot and the ULgrants are transmitted only in the second slot, as shown inFig. 14. This is intended to shorten the processing time toenable early PDSCH demodulation, which affects the im-plementation complexity.

Two R-PDCCH transmission methods are defined.- Without cross-interleaving: A single R-PDCCH occu-

pies dedicated frequency resource(s) and can support sin-gle rank beamforming by utilizing the DM-RS explainedin Sect. 2.2.

- With cross-interleaving: Several R-PDCCHs are inter-leaved in the frequency domain. This mode is suited to ob-taining the frequency diversity gain in the case that a rela-tively large number of RNs is associated with a d-eNB, orin the case that the d-eNB cannot accurately predict whichfrequency resources are most suitable for the R-PDCCHtransmission.

4.2.4 Hybrid ARQ Operation

For the access link, HARQ operation specified in Release 8is reused without any modification to retain backward com-patibility. Meanwhile, for the FDD system, MBSFN sub-frames cannot be configured in subframes #0,4,5, and 9 ina frame (10 ms) while the HARQ round trip time is 8 ms.Therefore, a backhaul link specific HARQ operation was in-troduced that retains the principle of Release 8 HARQ oper-ation as much as possible. For TDD, the HARQ round triptime is 10 ms so that no special handling is required.

4.3 Cooperative Operations for Relay Enhancement

The following part describes several coordination schemesfor relay enhancement. Some of these schemes were alsoconsidered in LTE-Advanced SI.

Network cording: The network coding scheme was pro-posed to conserve forwarding resources [24] and was pro-posed in LTE-Advanced SI. In the proposed network codingscheme, the downlink signal from the d-eNB and the uplinksignal from a UE are XOR-ed by signal to both the UE andthe d-eNB. However, it was not applicable to the LTE sys-tem due to the difference between the existing downlink anduplink frame formats, that is, the introduction of this scheme

would require a modification of one or both formats. Con-sequently, this approach cannot satisfy the backward com-patibility requirement.Resource management: To improve the overall system ca-pacity in relay systems, several papers have investigatedradio resource management schemes for backhaul and/oraccess links [25]–[27]. Time domain resource allocationsamong access and backhaul links for LTE-Advanced relaysystem were investigated in [25]. This paper shows basedon simulation results that the number of backhaul subframesthat achieves the maximum throughput is dependent on thenumber of RNs per d-eNB. Frequency domain resource par-titioning among access and backhaul links was studied in[26], [27]. These papers reveal that resource partitioningimproves both the system capacity and the cell edge userthroughput performance. However, for LTE-Advanced re-lay, frequency domain resource partitioning is only achievedin the transmission band domain, i.e., as an outband relay,to establish resource isolation between the backhaul link andthe access link.Interference coordination: Inter-cell interference (ICI) co-ordination is also an important technique to improve systemperformance. The resource partitioning can also be appliedfor ICI mitigation [28], [29]. These show that coordinatedresource partitioning can mitigate ICI among macro celland relay cells and that it leads to improvement in the sys-tem performance, especially for the cell edge user through-put. ICI mitigation based on backhaul subframe coordina-tion was also investigated [30].CoMP and eICIC application: LTE-Advanced relay wasdesigned with full L3 functionality and behaves as an eNBfrom the UE perspective. Therefore, CoMP and eICIC tech-niques described in the previous sections can also be appliedto the RN to improve the system capacity and the cell edgeuser throughput.

5. Conclusion

This paper presented an overview of radio interfacetechnologies for cooperative transmission in 3GPP LTE-Advanced, i.e., CoMP transmission, eICIC for hetero-geneous deployments, and relay transmission techniques.They are emerging techniques associated with enhanced net-work deployments that had not been considered when LTEinitial release (Release 8) was specified. In the 3GPP, stan-dardization efforts toward further enhancing the LTE radiointerface are continuing so that we can meet the future trafficdemand, which is rapidly growing.

References

[1] E. Dahlman, S. Parkvall, J. Skold, and P. Beming, 3G Evolution —HSPA and LTE for Mobile Broadband, 2nd ed., Academic Press,2008.

[2] S. Sesia, I. Toufik, and M. Baker, LTE — The UMTS Long TermEvolution from Theory to Practice, Wiley, 2009.

[3] 3GPP, TR36.814 (V9.0.0), “Further advancements for E-UTRAphysical layer aspects,” March 2010.


[4] M. Sawahashi, Y. Kishiyama, A. Morimoto, D. Nishikawa, and M.Tanno, “Coordinated multipoint transmission/reception techniquesfor LTE-Advanced,” IEEE Trans. Wirel. Commun. Mag., vol.17,no.3, pp.26–34, June 2010.

[5] S. Jing, D.N.C. Tse, J.B. Soriaga, J. Hou, J.E. Smee, and R.Padovani, “Multicell downlink capacity with coordinated process-ing,” EURASIP J. Wirel. Commun. Netw., vol.2008, pp.1–19, 2008.

[6] K. Sakaguchi, “Base station cooperation MIMO for spatial spectrumsharing,” IEICE Technical Report, SR2007-103, March 2008.

[7] S.A. Ramprashad, H.C. Papadopoulos, A. Benjebbour, Y.Kishiyama, N. Jindal, and G. Caire, “Cooperative cellular networksusing multi-user MIMO: Tradeoffs, overheads, and interferencecontrol across architectures,” IEEE Commun. Mag., vol.49, no.5,pp.70–77, May 2011.

[8] D. Love, R. Heath, V. Lau, D. Gesbert, B. Rao, and M. Andrews,“An overview of limited feedback in wireless communication sys-tems,” IEEE J. Sel. Areas Commun., vol.26, no.8, pp.1341–1365,Oct. 2008.

[9] V.R. Cadambe and S.A. Jafar, “Interference alignment and degreesof freedom of the K-user interference channel,” IEEE Trans. Inf.Theory, vol.54, no.8, pp.3425–3441, Aug. 2008.

[10] 3GPP, R1-110576, LG Electronics, Motorola Mobility, Intel, Mar-vell, Samsung, ETRI, Orange, Renesas, Huawei, “Proposal forUE receiver assumption in CoMP simulations,” TSG-RAN WG1#63bis, Jan. 2011.

[11] 3GPP, R1-110461, Ericsson, ST-Ericsson, “Baseline schemes andfocus of CoMP studies,” TSG-RAN WG1 #63bis, Jan. 2011.

[12] Qualcomm, “LTE Advanced: Heterogeneous networks,” URL:http://www.qualcomm.com/

[13] A. Khandekar, N. Bhushan, J. Tingfang, and V. Vanghi, “LTE-Advanced: Heterogeneous networks,” 2010 European Wireless Con-ference, pp.978–982, April 2010.

[14] 3GPP, R1-105081, CMCC, “Summary of the description of candi-date eICIC solutions,” Aug. 2010.

[15] 3GPP, R1-104463, Nokia Siemens Networks, “Autonomous powersetting for HeNB cells,” Aug. 2010.

[16] M. Morita, Y. Matsunaga, and K. Hamabe, “Adaptive power levelsetting of femtocell base stations for mitigating interference withmacrocells,” IEEE VTC 2010-Fall, pp.1–5, 2010.

[17] 3GPP, R1-104415, Alucatel-Lucent, “Performance of power controlmethods in macro-femto deployment,” Aug. 2010.

[18] A.B. Saleh, S. Redana, B. Raaf, T. Riihonen, J. Hamalainen, and R.Wichman, “Performance of amplify-and-forward and decode-and-forward relays in LTE-Advanced,” IEEE VTC 2009-Fall, pp.1–5,2009.

[19] 3GPP TS36.106, “FDD Repeater radio transmission and reception.”[20] W. Zhang and U. Mitra, “Channel-adaptive frequency-domain relay

processing in multicarrier multihop transmission,” IEEE ICASSP2008, pp.3229–3232, March 2008.

[21] J.N. Laneman and G.W. Wornell, “Distributed space-time codedprotocols for exploiting cooperative diversity in wireless networks,”IEEE Trans. Inf. Theory, vol.49, no.10, pp.2415–2425, Oct. 2003.

[22] 3GPP TS36.216, “E-UTRA physical layer for relaying operation.”[23] 3GPP TS23.401, “GPRS enhancements for E-UTRAN access.”[24] R. Ahlswede, N. Cai, S.-Y.R. Li, and R.W. Yeung, “Network infor-

mation flow,” IEEE Trans. Inf. Theory, vol.46, no.4, pp.1204–1216,July 2000.

[25] S. Roth, G. Jiansong, and D. Danev, “Subframe allocation for re-lay networks in the LTE Advanced standard,” IEEE PIMRC 2010,pp.1758–1763, Sept. 2010.

[26] T. Liu, M. Rong, P. Li, D. Yu, Y. Xue, and E. Schulz, “Radio re-source allocation in two-hop cellular relaying network,” IEEE VTC2006-Spring, pp.91–94, May 2006.

[27] L. Huang, M. Rong, L. Wang, Y. Xue, and E. Schulz, “Resource al-location for OFDMA based relay enhanced cellular networks,” IEEEVTC 2007-Spring, pp.3160–3164, April 2007.

[28] M. Liang, F. Liu, Z. Chen, Y. Wang, and D.C. Yang, “A novel fre-

quency reuse scheme for OFDMA based relay enhanced cellular net-works,” IEEE VTC2009-Spring, pp.1–5, April 2009.

[29] J. Liu, D. Wang, J. Pang, J. Wang, and G. Shen, “Inter-cell interfer-ence coordination based on soft frequency reuse for Relay enhancedcellular network,” IEEE PIMRC 2010, pp.2304–2308, Sept. 2010.

[30] Y. Yuda, A. Iwata, and D. Imamura, “Interference mitigation us-ing coordinated backhaul timing allocation for LTE-Advanced relaysystems,” IEEE ICC2011, June 2011.

Tetsushi Abe received B.S. degree and M.S.degree in Electrical and Electronic Engineering,and Dr. Eng. Degree in Communications and In-tegrated Systems from Tokyo Institute of Tech-nology, Tokyo, Japan in 1998 and 2000 and2007 respectively. During 1998–1999 he stud-ied in Electrical and Computer engineering inUniversity of Wisconsin, Madison, U.S.A. un-der the scholarship exchange student programoffered by the Japanese Ministry of Education.He joined NTT DOCOMO in 2000. During

2005–2009, he was with the DOCOMO Euro-Labs, Munich, Germany.Since August 2009, he has been serving as a vice chairman of 3GPP RANWG1. His current research interests include wireless communication tech-nologies for 3GPP standards.

Yoshihisa Kishiyama received his B.E.,M.E., and Dr. Eng. degrees from Hokkaido Uni-versity, Sapporo, Japan in 1998, 2000, and 2010,respectively. In 2000, he joined NTT DO-COMO, INC. His research interests include ra-dio access technologies for 3G long-term evolu-tion and for the systems beyond IMT-Advanced.He is a member of the IEEE.

Yoshikazu Kakura received the B.E. andM.E. degrees in computer science and commu-nication engineering from Kyusyu University in1990 and 1992, respectively. He joined NECCorporation in 1992. He is engaged in researchfor beyond 3G mobile communication system.

Daichi Imamura received the B.E. de-gree from Kanazawa University, Japan in 1995and received the M.E. degree from Osaka Uni-versity, Osaka Pref., Japan in 1998. He joinedMatsushita Communication Ind. Co., Ltd.,Japan in 1998. He has been with Panasonic Cor-poration, Japan since 2004. He is engaged inthe research and development of mobile com-munication systems and standardization. He is amember of the IEEE.

radio interface technologies for cooperative transmission

Documents